search

alex-schaaf / abc-geomodeling

Figuring out how Approximate Bayesian Computation works and how it can be applied to geological modeling.
2 stars 0 forks source link

Error in finding interp_data.geo_data_res #1

Open Ali1990dashti opened 4 years ago

Ali1990dashti commented 4 years ago

Dear @alex-schaaf

hope doing well. Firstly I appreciate your informative examples on Gempy. Today I was trying to run one of your examples (https://github.com/alex-schaaf/ABC/blob/master/gempy_abc_topo.ipynb) but I faced an error after trying to copy interp_data.geo_data_res. I am using the last released version of Gempy (2.2.2). I copied the screenshot of the error here: Fehler In advance I appreciate any help and response. Regards, Ali

alex-schaaf commented 4 years ago

Hey @Ali1990dashti !

I didn't expect anyone ever to look at my exploratory notebooks for coupling gempy with approximate bayesian computation methods :smile: Thus I sadly did not upload a requirements or environment file with the proper module verions used :disappointed:

So the notebook has not been updated sice Latest commit a528d50 on Jun 21, 2018 - so the error is due to differences in gempy versions. If I recall correctly this notebook / repository was still written with GemPy 1.x. So you could try installing a version of GemPy released around June 2018 - maybe Gempy 1.14 https://github.com/cgre-aachen/gempy/releases/tag/v1.14

But note that GemPy 1 and 2 are quite different. It might be easier to look at the official GemPy probabilistic modeling tutorial https://docs.gempy.org/tutorials/ch5_probabilistic_modeling/ch5_1.html for Monte Carlo error propagation and add your own ABC-REJ logic to it :smile:

Ali1990dashti commented 4 years ago

Dear @alex-schaaf

Firstly, thanks for providing such a detailed answer to me. I truly apprciate it. I also visited both the gempy.org and also the informative github page of the gempy. Can I ask what _interp_data.geo_datares is? Is it a part of the theano interpolator? When I aks about its type, python says it's a gempy.core.interpolator.InterpolatorModel. Again, thanks for giving time to me.

Cheers, Ali

alex-schaaf commented 4 years ago

The input data for GemPy models is being rescaled between 0 and 1 for the computation in theano, and this old notebook uses the dataframe with the rescaled input data to modify this data directly for the stochastic simulation.

Ali1990dashti commented 4 years ago

Sorry for making that much questions and taking your time ... I know that input data are rescaled in Gempy using rescaling factor and centers. But the confusing point for me is that in the notebook, the rescaled data are called from the interp_data which has been the result of theano compilation (cell number 71). Does it mean that all the input data are stored as interp data in theano? Again sorry for making that much questions ...